Figure 1
From: Identifying pathogenic processes by integrating microarray data with prior knowledge
Results on simulated data. We have used the simulation scheme proposed by [26] sampling approximately 100 genes per data set. Sample sizes (N) were varied using N = 10, 100 and 1000, as well as an extra variation (SD) added to each element in the expression matrix, using SD = 0, 1 and 2. hclust = hierarchical clustering, kmeans = k - means clustering, PAM = Prediction Around Medoids, Mclust = model-based clustering, tight = tight clustering, MCIP-A, is our method (MCMC Clustering using Informative Priors), but with no priors used, MCIP-B is our method using priors with 20% of the priors mis-specified, and MCIP-C is our method with all prior pairs correctly specified.